A Fast Global Motion Estimation for Moving Objects Segmentation Using Moment-Preserving Technique
Shyi‐Chyi Cheng, Wei-Kan Huang, Tian-Luu Wu · 2005
This paper presents a fast segmentation of moving objects in video sequences using the moment-preserving technique. The emerging video coding standard MPEG-4 enables various content-based functionalities for multimedia applications. To decompose each frame of a video sequence into video object planes (VOP) is the first step for MPEG-4 toward supporting such functionalities, as well as improving coding efficiency. Each VOP corresponds to a single moving object in the scene. For this purpose, we propose a novel region-based segmentation method as a general VOP segmentation algorithm by estimating the motion vector of each region in a VOP using the proposed moment-preserving technique. An initial spatial partition of each frame is obtained by a fast, region-growing image segmentation algorithm. The solution to the motion vector of each region in a VOP is analytic. This algorithm can be performed very fast for content-based multimedia applications with no need for special hardware. Experimental results for several video sequences demonstrate the effectiveness of the proposed approach.